well there is https://artificialanalysis.ai/evaluations/omniscience
It's a gibberish input detection benchmark, and does not measure output hallucinations.
alt Hacker News
It's a gibberish input detection benchmark, and does not measure output hallucinations.